Adult attachment and prolonged grief

Diagnoses characterized by severe, persistent and disabling grief have recently been added to the ICD-11 and DSM-5-TR as prolonged grief disorder. Adult attachment is widely assumed critical in the development, persistence, and treatment of prolonged grief, yet a meta-analysis on this topic is lacking. We conducted a systematic review (PROSPERO: CRD42021220511) searching PsycInfo, Web of Science, and PubMed (final search: August 2022) to identify and summarize quantitative research examining relationships between adult attachment (i.e., attachment anxiety, attachment avoidance, secure attachment, disorganized attachment) and prolonged grief symptoms. Thirty-one studies including 8347 bereaved adults were included. Attachment anxiety ( r = 0.28, 95 % CI:0.23 – 0.32, k = 15) and attachment avoidance ( r = 0.15, 95 % CI:0.05 – 0.26, k = 15) related positively to prolonged grief symptoms concurrently. We found no evidence of publication bias but did detect heterogeneity in effect sizes. Ten longitudinal analyses showed no evidence that insecure attachment styles increase prolonged grief symptoms. Attachment anxiety predicted better therapy outcomes. Insecure attachment styles are concurrently positively related to prolonged grief symptoms but do not increase grief severity. The role of adult attachment in contemporary grief theories may need reconsideration. Intensive longitudinal research should aim to clarify how dynamic changes in attachment to the deceased and others relate to changes in prolonged grief symptoms.


Prolonged grief
There is increasing recognition that a minority of bereaved individuals experience severe, persistent and impairing grief, termed prolonged grief.In 2018, the International Classification of Diseases eleventh edition (ICD-11, World Health Organization, 2022) included a diagnosis characterized by prolonged grief, termed prolonged grief disorder (PGD).PGD is characterized by pervasive yearning for the deceased and/or cognitive preoccupation with the deceased, combined with one or more symptoms indicative of emotional pain.A related but distinct version of PGD has recently been included in the text revision of the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5-TR: American Psychiatric Association; Eisma, 2023).While effective interventions for prolonged grief have been developed (e.g., Boelen et al., 2007;Bryant et al., 2014;Johannsen et al., 2019;Shear et al., 2005), no more than half of all clients experience clinically relevant changes following such treatments (Doering & Eisma, 2016).Therefore, it appears worthwhile to improve understanding of factors theorized to be involved in the etiology of prolonged grief as such factors provide the basis for current treatments for prolonged grief.

Attachment theory
Since the seminal work of John Bowlby (1980), attachment, a biopsychosocial motivational system affecting how we relate to others, has been firmly on the map as a potential determinant of bereavement outcomes.It is a cornerstone to many prominent theories on coping with bereavement and prolonged grief (e.g., LeRoy et al., 2019;Maccallum & Bryant, 2013;Shear & Shair, 2005;Stroebe et al., 2005) and forms the basis for psychological treatments, e.g., attachment-informed grief therapy (Kosminsky & Jordan, 2016).In brief, attachment theory holds that an infant's relationship with their primary caregiver influences their attachment relationships later in life (Bowlby, 1973).Originally, three distinct attachment styles were distinguished: secure, anxious, and avoidant (Ainsworth et al., 1978) and disorganized attachment was later added to this list (Main & Solomon, 1990).These styles were proposed to translate into two different attachment dimensions: attachment avoidance and attachment anxiety (Fraley & Shaver, 2000).A person with an avoidant attachment style (i.e., high on attachment avoidance) would typically tend to feel uneasy with closeness, maintain emotional distance, and strive for excessive independence.A person with an anxious attachment style (i.e., high on attachment anxiety) would typically be preoccupied with intimate relationships, fear rejection, and worry about abandonment.A person with a disorganized attachment style, scoring high on both dimensions, will view others as unpredictable and feel unsure whether he or she will be supported by them.A securely attached person scores low on both dimensions, generally keeping a healthy balance between dependence and independence (Fraley & Shaver, 2000).

Attachment theory and (prolonged) grief
Following the loss of a loved one, people with different attachment styles are assumed to respond in diverging ways (Shaver & Tancredy, 2001).At the risk of oversimplification: securely attached individuals will react emotionally to loss, but will not feel emotionally overwhelmed by grief.They create a coherent narrative about the loss and its consequences and do not suffer intense self-blame or lowered self-esteem.Insecurely attached individuals will respond in less functional ways.Avoidantly attached individuals will likely not become very emotional about loss, because they generally depend less on their loved ones emotionally, which may elicit delayed or inhibited grief.Anxiously attached individuals will typically respond to loss with stronger emotional responses and preoccupation with the lost relationship.Individuals characterized by disorganized attachment are likely to be confused about their loss-related emotions and could find it difficult to create a coherent narrative about the loss (Shaver & Tancredy, 2001).
In line with these theoretical notions, insecure attachment styles are proposed to be a risk factor of severe and persistent grief reactions across a wide range of prominent (prolonged) grief models (e.g., Shear & Shair, 2005;Stroebe et al., 2005;Maccallum & Bryant, 2013;Mikulincer & Shaver, 2022), which form the basis for bereavement care (e.g., Kosminsky & Jordan, 2016;Newsom et al., 2017;Shear et al., 2016).Across grief models, insecure attachment is presumed to contribute the persistence of severe grief reactions, via different proposed mechanisms.For example, within Stroebe et al.'s (2005) Dual Process Model, flexible oscillation between coping with loss-related stressors (stressors relating directly to the loss) and restoration-oriented stressors (stressors that come about as secondary consequences of the loss) is necessary for healthy adaptation to bereavement.An anxious attachment style is proposed to result in a heightened focus on the loss, an inability to cope with the loss alone and, in extreme cases, a persistent, severe grief reaction accompanied by rumination and preoccupation with the lost relationship (i.e., prolonged grief).An avoidant attachment style, on the other hand, is thought to be more likely to result in suppression of grief and a strong focus on restoration-oriented stressors.In extreme cases, avoidantly attached persons might show inhibited, absent, or delayed grief.
Similarly, Mikulincer and Shaver (2022) posit that loss-related distress is affected by a person's attachment orientation.Attachment anxiety, being associated with a personal sense of vulnerability and helplessness as well as overdependence on relationship partners for coping with threats and challenges, is proposed to intensify emotional reactions to interpersonal loss.Therefore, the loss of a close relationship can seem catastrophic for prototypically anxiously attached individuals.Conversely, people shoring high on attachment avoidance are proposed to be less affected by losing an attachment figure, partially because they generally downplay their attachment needs and rely on themselves for coping with losses.However, in some circumstances, the vulnerability of an avoidantly-attached individual may become apparent over time, for instance when suppression and ignoring fail to manage distress (Mikulincer & Shaver, 2022).

The current review
In view of the theoretical and clinical importance of adult attachment in prolonged grief, understanding this relationship appears critically important.While prior narrative reviews and one systematic review on adult attachment and prolonged grief have been published (e. g., Maccallum & Bryant, 2013;Mikulincer & Shaver, 2008, 2022;Russ et al., in press;Schenck et al., 2016), a comprehensive systematic review with a meta-analysis is lacking.While narrative reviews can be useful for theoretical development, systematic reviews typically provide higher quality answers to specific empirical questions (Mallett et al., 2012).Systematic reviews generally consider a larger body of literature, while answering pre-determined questions.They focus on empirical evidence instead of preconceived knowledge, and follow pre-registered rules of evidence synthesis, thereby improving transparency and replicability.Additionally, meta-analyses provide a helpful method of evaluating and synthesizing quantitative data, by enhancing the accuracy of effect size estimates through increases of the number of observations and statistical power (Fagard et al., 1996).
In summary, the main objective of the present review is to clarify the nature of the relationship between adult attachment and prolonged grief symptoms.Specifically, we aimed to synthesize knowledge about the associations between attachment styles and prolonged grief symptoms, derived from quantitative empirical studies published since the emergence of validated instruments to assess prolonged grief symptoms.Ultimately, our aim was to provide a comprehensive overview of the knowledge on this topic to enable a critical evaluation of attachmentbased theories of prolonged grief, thereby providing guidance for future research and the improvement of treatments for severely distressed bereaved persons.

Preregistration
This systematic review and meta-analysis were preregistered in PROSPERO's international registry of systematic reviews under registration number CRD42021220511.

Search strategy
The first author (ME), an experienced systematic reviewer, searched the databases PsycInfo, Web of Science, and PubMed.We used the following keywords: "prolonged grief" OR "complicated grief" OR "persistent complex bereavement-related disorder" OR "traumatic grief" OR "disturbed grief" OR "pathological grief" AND "attachment" OR "attachment style" OR "attachment theory" OR "insecure attachment" OR "secure attachment" OR "anxious attachment" OR "avoidant attachment" OR "disorganized attachment" OR "preoccupied dependent attachment" OR "avoidant dismissive attachment".We limited our searches to work published after 1994, as the first instrument to assess prolonged grief symptoms emerged in 1995 (Inventory of Complicated Grief: Prigerson et al., 1995).We conducted our final search on the 23rd of August 2022 including 618 potentially relevant papers (across our M.C.Eisma et al. original and final searches).Subsequently, 205 duplicates were removed, leaving 413 papers for screening.Two Master students (KB and LA) carried out screening independently in Covidence.In case of disagreement, the two authors discussed until consensus was reached.A third researcher (ME) provided a final decision in case there was still disagreement after discussion.After the screening of title and abstracts, 85 papers were selected.After full-text screening, 29 articles remained.One eligible paper was added after inspection of the reference list of included studies and two after inspection of a previous systematic review (Russ et al., in press), leaving 32 papers (reporting on 31 studies) to review.Fig. 1 shows a PRISMA flowchart of the search.

Inclusion and exclusion criteria
As mentioned, we included only papers published after 1994 because no validated measures of prolonged grief symptoms were available before this time.To safeguard interpretability and study quality, only quantitative, peer-reviewed scientific articles, written in English were included.For ease of interpretation, we included only articles reporting research on adult samples who experienced the death of another human being.Moreover, to ensure adequate power for detection of a very large correlation between the investigated constructs (r = 0.80, Cohen, 1988), we only included studies with a minimum sample size of 20 bereaved adults.Additionally, included studies had to report a statistical relationship between an attachment measure and a prolonged grief symptom measure.
We excluded papers reporting exclusively on constructs related to but dissimilar from attachment (e.g., continuing bonds).Articles were excluded if they used measures of grief showing limited content overlap with contemporary prolonged grief symptom measures (e.g., Texas Revised Inventory of Grief, TRIG, Faschingbauer, 1981) and/or exclusively measures of other mental health problems (e.g., posttraumatic stress disorder, depression, general distress).Similarly, to allow for straightforward interpretations we excluded studies combining measures of prolonged grief symptoms with measures of normal grief reactions or symptoms of other mental health problems (e.g., Mancini et al., 2009).Lastly, papers reporting on the same data and the same associations between adult attachment and prolonged grief symptoms were included only once (with exception of two reports on one dataset reporting different sample sizes: Meert et al., 2010Meert et al., , 2011, which will both be discussed but are counted as one study).

Data extraction
The following data was extracted: first author, year of publication, country of research, sample size, demographic characteristics (i.e., age, gender), loss-related characteristics (i.e., time since loss, relationship with the deceased, cause of death), study design, characteristics of the attachment measure and prolonged grief symptom measure (i.e., name, number of items, mean score), the statistical associations between attachment and prolonged grief symptoms, and the effect sizes of the associations (see Table 1).Regarding the extraction of the statistical effects it should be noted that for studies reporting on cross-sectional surveys we exclusively extracted zero-order correlations (if available).Zero-order correlations provide information about simple linear associations without correction for effects of other variables, facilitating straightforward interpretation of findings.
We conducted a brief study quality assessment for studies with dissimilar designs, in line with Eisma and Stroebe (2021), who rated research according to six criteria.First, we established whether studies provided a comprehensive description of sample characteristics (i.e., included the following information: gender, age, time since loss, relationship with the deceased, cause of death, or all of the previous characteristics except one; 1 point) vs. non-comprehensive description (i.e., two or more sample characteristics missing; 0 point).Next, we rated study design (longitudinal survey, 1 point vs. cross-sectional survey, 0 point); inclusion of a control group (yes, 1 point vs. no, 0 point); use of validated measures for attachment (yes, 1 point vs. no, 0 point); and use        of validated measures for prolonged grief symptoms (yes, 1 point vs. no, 0 point).Lastly, we established if effect sizes were reported (yes, 1 point vs. no, 0 point).Study quality scores can range from zero to six, with higher scores indicating higher study quality.Two Master students (KB and LA) discussed discrepancies between extracted data until they achieved consensus.If no consensus could be reached, a third researcher (ME) was consulted and made a final decision.Initial interrater agreement between the two researchers across all extracted data was 93 %.Lastly, one experienced systematic reviewer (BD) double-checked all extracted data to ensure data quality.

Meta-analyses
Longitudinal analyses were too diverse to allow for meta-analyses.Therefore, meta-analyses were limited to studies that reported crosssectional zero-order correlations (including baseline associations in longitudinal studies) for measures of adult attachment and prolonged grief symptoms (k = 15) and were conducted separately for attachment anxiety and attachment avoidance.Pearson's product-moment correlation coefficient (r) was used as the measure of effect size and interpreted in line with common standards for effect sizes (r = 0.10 small; r = 0.30 moderate; r = 0.50 large ;Cohen, 1988).Where studies reported measures of attachment in general and attachment to the deceased in the same sample simultaneously, only measures of attachment in general were included in order to ensure independence of effect sizes (k = 1; Jerga et al., 2011).All analyses were conducted using Excel with the add-on software package MetaXL v5.3 (Barendregt & Doi, 2016) and used random effects models.Doi plot asymmetry was investigated using the Luis Furuya-Kanamori (LFK) index.The LFK index tests for possible publication bias with an index between 0 and ±1 indicating no asymmetry, between ±1 and ±2 indicating minor asymmetry, and >±2 indicating major asymmetry.Heterogeneity among studies was inspected using Cochran's Q-test and the I 2 statistic.We considered substantial heterogeneity present if the Q-test was significant and I 2 exceeded 50 %.Since the attachment measures differed across studies, we planned to run subgroup analyses comparing effects for studies employing different attachment measures (attachment in general vs. attachment to the deceased).Outliers were defined as studies with a 95 % CI for their effect size which did not overlap with the 95 % CI for the overall pooled effect size.All analyses were recalculated after exclusion of outliers.Additionally, we conducted sensitivity analyses with MetaXL to identify outliers.

Study characteristics
Table 1 contains study descriptions.The 32 identified papers reported 31 studies on 8347 bereaved participants and 349 non-bereaved participants (excluding overlapping datasets).Twenty-one studies reported on cross-sectional surveys (68 %) and 10 on longitudinal surveys (32 %).Twenty-one studies (68 %) employed a version of the Inventory of Complicated Grief (ICG: Prigerson et al., 1995), and eight studies (26 %) employed a version of the Prolonged Grief-13 (PG-13; Prigerson et al., 2009), as a measure of prolonged grief symptoms.One study based its measure on the criteria for Persistent Complex Bereavement Disorder (PCBD, DSM-5: American Psychiatric Association, 2013).Finally, one study derived symptoms of severe grief from previous grief studies (e.g., Horowitz et al., 1997).Summary of the main (unweighted) sample characteristics across studies were as follows: Bereaved participants were on average 42.23 years old (SD = 12.32) and predominantly female (75 %).Sixty-eight percent (n = 21) of the investigations included samples of people with varying relations to the deceased.Those focusing on specific relationships to the deceased most frequently studied spousal loss (13 %; n = 4), or child loss (13 %, n = 4).Cause of death was classified into nonviolent versus violent (i.e., accidents, suicide, homicide).Across 21 studies that included information on these causes of death, a minority of bereaved adults experienced violent deaths (39 %).Five studies exclusively included people bereaved due to violent causes (24 %).Twenty-four studies reported the mean time since loss which was on average 34.9 months (SD = 34.2).
Thirty studies used a dimensional approach to assess the attachment anxiety and attachment avoidance.Only one study took a categorical approach distinguishing four main categories of attachment as identified by Ainsworth et al. (1978) and Main and Solomon (1990) (secure, anxious, avoidant, disorganized).Twenty-five studies measured attachment within close relationships in general (81 %).Five studies assessed attachment to the deceased (16 %).One study measured both types of attachment.All studies included measures of attachment anxiety and attachment avoidance.Only three studies included measures specifically assessing secure attachment (10 %) and four studies assessed disorganized attachment (13 %).characteristics were described comprehensively in 28 papers (90 %).

Study quality
Ten (32 %) papers reported longitudinal surveys.Twenty-one (68 %) were cross-sectional surveys.One study included a non-bereaved control group (3 %) (Pini et al., 2012).No studies included unvalidated measures for attachment styles yet two studies used an unvalidated measure for prolonged grief symptoms (6 %) (Fraley & Bonanno, 2004;van der Houwen et al., 2010).Effect sizes were reported in 21 papers (68 %).In the presentation of main findings, we will mention design as well as reported effect sizes for the association between adult attachment and prolonged grief symptoms.For the concurrent associations, we report zero-order correlations, when available.

Meta-analysis of concurrent associations between attachment anxiety and prolonged grief symptoms
The random effects model demonstrated a pooled correlation coefficient of small size between attachment anxiety and prolonged grief symptoms (r = 0.28; k = 15; 95 % CI: 0.23-0.32,Fig. 2).The LFK index was − 0.78, indicating no asymmetry (i.e., no publication bias).Heterogeneity among studies was significant (Q = 34.68,p ≤ .001;I 2 = 60 %).The subgroup analysis comparing studies that measured attachment anxiety in general (r = 0.29; k = 11; 95 % CI: 0.24-0.33;I 2 = 52 %) and studies that measured attachment anxiety with regard to the deceased (r = 0.25; k = 4; 95 % CI: 0.12-0.37;I 2 = 78 %) found no significant difference between the groups, Q(1) = 0.47, p = .493.Table 3 presents the results of the MetaXL sensitivity analysis showing the pooled correlation coefficient when each individual respective study would be excluded.No exclusion of any single study led to a pooled correlation coefficient that was outside of the initial pooled correlation coefficient Note.PG = prolonged grief.Burke et al. (2019) is counted as a cross-sectional study, because it is unclear if reported associations between attachment styles and PG are longitudinal.Meert et al. (2010Meert et al. ( , 2011) ) is counted as a longitudinal study.
confidence interval.Thus, the sensitivity analysis identified no outliers.

Qualitative description of concurrent associations between secure attachment and disorganized attachment with prolonged grief symptoms
Two studies showed a significant negative relationship between attachment security and prolonged grief symptoms (Levi-Belz & Lev-Ari, 2019;Pini et al., 2012).One study with a small sample size found a nonsignificant result for the relationship between secure attachment and prolonged grief symptoms (Joyce et al., 2010).The same studies examined the relationship between disorganized attachment and prolonged grief symptoms, which was non-significant in all three papers.However, one recent study reported a positive correlation between disorganized attachment and prolonged grief symptoms (Sekowski & Prigerson, 2022).Due to the low number of studies, no meta-analyses were conducted for the associations between these attachment styles and prolonged grief symptoms.

Qualitative description of longitudinal associations between attachment anxiety and prolonged grief symptoms
Longitudinal associations between attachment anxiety and prolonged grief symptoms were reported in ten studies.Three of these studies reported regression analyses not controlling for baseline prolonged grief symptoms, showing small and moderate positive associations between attachment anxiety at baseline and prolonged grief symptoms at follow-up (Harris et al., 2021;Meert et al., 2011;Milman et al., 2019).Four studies reported regression analyses controlling for baseline symptoms, yielding no significant associations between attachment anxiety at baseline and prolonged grief symptoms at followup (Boelen, 2012;Harris et al., 2021;Lee et al., 2020;Suttle et al., 2022).Another two studies examined the association of attachment anxiety with prolonged grief symptom change, reporting no significant associations (Fraley & Bonanno, 2004;Meert et al., 2011).Moreover, two studies employed longitudinal multilevel models investigating the interaction of attachment anxiety and time on prolonged grief symptoms.Van der Houwen et al. (2010) ran a model in which time (level 2) was nested in individuals (level 1), showing that attachment anxiety did not interact with time to predict prolonged grief symptoms.Wijngaardsde Meij et al. (2007a) ran a model in which time (level 3) was nested in individuals (level 2), which were nested in couples (level 1), showing that an individual's attachment anxiety interacted with time, so that more anxiously attached persons showed a greater decline (i.e., faster recovery) in prolonged grief symptoms over time.One recent study similarly showed that attachment anxiety related to stronger prolonged grief symptom reductions during internet-based cognitive-behavior therapy (Schmidt et al., 2022).

Qualitative description of longitudinal associations between attachment avoidance and prolonged grief symptoms
Longitudinal associations between attachment avoidance and prolonged grief symptoms were also reported in ten studies.Of three studies reporting analyses not controlling for baseline prolonged grief symptoms, two showed small and moderate positive associations between avoidant attachment at baseline and prolonged grief symptoms at follow-up (Meert et al., 2011;Milman et al., 2019) while one showed null-results (Harris et al., 2021).Four studies reported analyses controlling for baseline symptoms, finding no significant associations between avoidant attachment at baseline and prolonged grief symptoms at follow-up (Boelen, 2012;Harris et al., 2021;Lee et al., 2020;Suttle et al., 2022).Another two studies examined the association of attachment avoidance with PG symptom change, reporting no significant associations (Fraley & Bonanno, 2004;Meert et al., 2011).Furthermore, two studies employed longitudinal multilevel models investigating the interaction of attachment avoidance and time on prolonged grief symptoms.Van der Houwen et al. ( 2010) ran a model in which time (level 2) was nested in individuals (level 1), showing that attachment avoidance did not interact with time to predict prolonged grief symptoms.Wijngaards-de Meij et al. (2007) ran a model in which time (level 3) was nested in individuals (level 2), which were nested in couples (level 1), showing that attachment avoidance did not interact with time to predict prolonged grief symptoms.One study showed that attachment avoidance did not predict symptom change during internet-based cognitive-behavioral therapy (Schmidt et al., 2022).

Discussion
The aim of this systematic review was to provide a comprehensive overview of the associations between adult attachment and prolonged grief symptoms.Turning to cross-sectional surveys first, most surveys indicated that small to moderate positive associations exist between attachment anxiety and prolonged grief symptoms and between attachment avoidance and prolonged grief symptoms.However, some studies reported null results and (in the case of attachment avoidance) negative associations.Meta-analyses showed that on average small associations exist between attachment anxiety and attachment avoidance, on the one hand, and prolonged grief symptoms, on the other hand.We found no evidence of reporting bias, but did find heterogeneity in effects.For the relationship between attachment avoidance and prolonged grief symptoms, this effect was partially explained by the type of attachment relationship under investigation.Attachment avoidance in the relationship to the deceased was not significantly associated with prolonged grief symptoms, whereas general attachment avoidance was related to prolonged grief symptoms.A negative concurrent association between secure attachment and prolonged grief symptoms was found in two of three studies reporting on it.Disorganized attachment did not relate significantly to prolonged grief symptoms in three studies but one study did report positive associations between these constructs.
Turning to longitudinal surveys next, we found evidence of small to moderate positive associations of attachment anxiety with prolonged grief symptoms in three longitudinal analyses that did not take into account autocorrelations of prolonged grief symptoms (Harris et al., 2021;Meert et al., 2011;Milman et al., 2019).For the association of attachment avoidance and prolonged grief symptoms a similar picture emerged for two studies (Meert et al., 2011;Milman et al., 2019), but there was also one null-result (Harris et al., 2021).In stringent longitudinal analyses (i.e., those controlling for baseline symptoms, analyzing change scores, or assessing the interaction between time and attachment styles on prolonged grief symptoms within multilevel models) attachment anxiety and attachment avoidance did not increase prolonged grief symptoms (Boelen, 2012;Fraley & Bonanno, 2004;Harris et al., 2021;Lee et al., 2020;Meert et al., 2011;Suttle et al., 2022;van der Houwen et al., 2010).One study reported that attachment Note: LCI = lower bound confidence interval, HCI = higher bound confidence interval.
M.C.Eisma et al. anxiety levels predicted decreases of prolonged grief symptoms over time (Wijngaards-de Meij et al., 2007a).Therefore, we found no convincing evidence that insecure attachment fuels prolonged grief symptoms.Since temporal precedence of effects is an important prerequisite of demonstrating causal relationships (van den Hout et al., 2017), the present evidentiary base provides no clear support for causal relationships between adult attachment and prolonged grief symptoms.Interestingly, one study showed that people with higher attachment anxiety benefited more from cognitive behavioral therapy for prolonged grief (Schmidt et al., 2022), suggesting the potential clinical relevance of tailoring treatments to attachment styles.Indeed, Kosminsky and Jordan (2016) have argued that the grief process of insecurely attached bereaved adults may be more complicated and treatments should be nuanced and tailored to their attachment orientation.An attuned therapist can help a client regulate and begin to focus flexibly, openly and honestly, on the work of establishing a continued bond with the deceased while attending to the demands of their new life that remain a part of it.
Taken together, the results from this systematic review confirm that attachment insecurity relates to prolonged grief, but contradict the central premise of theories of (prolonged) grief suggesting that attachment insecurity plays a causal role in the persistence of severe grief reactions (e.g., Maccallum & Bryant, 2013;Shaver & Tancredy, 2001;Shear & Shair, 2005;Stroebe et al., 2005).Nevertheless, attachment anxiety appears important to consider as a determinant of the effects of grief therapies (Schmidt et al., 2022).What do these mixed findings tell us about the usefulness of attachment theory in understanding adaptation to bereavement?A number of relevant considerations regarding the interpretation of findings and avenues for future research are discussed next.
First, we should consider to what extent we can expect one's attachment orientation, or general attachment style (assessed in most reviewed studies), to relate meaningfully to prolonged grief after the loss of a loved one with which one has had a specific attachment relationship.The nature of attachment bonds as well as the degree of attachment may vary from one personal relationship to the next (Fraley et al., 2011;Shaver & Tancredy, 2001).Indeed, some contemporary attachmentbased models of coping with interpersonal loss hold that it is not one's attachment style per se, but rather the dynamic changes in attachment to a lost person that (co)determine bereavement outcome (e.g., LeRoy et al., 2019;Mikulincer & Shaver, 2022;Shear & Shair, 2005).For example, LeRoy et al. (2019) state that people in one's social environment fulfill attachment related functions.Bowlby (1982) distinguished the functions of proximity seeking (seeking closeness of an attachment figure and resisting separation from this person), safe haven (an attachment figure reduces distress and offers support during difficulties) and secure base (an attachment figure provides a core sense of emotional and psychological security) (see: Hazan et al., 2004).Among adults, the partner is typically at the top of the attachment hierarchy, and fulfills most attachment-related functions, whereas secondary attachment figures, such as parents, siblings and/or close friends are on lower levels of the attachment hierarchy (Trinke & Bartholomew, 1997).Following the loss of an attachment figure, the extent to which one succeeds in reorganizing one's attachment hierarchy to fulfill the attachment-related functions without the lost person, may determine physiological and psychological recovery.Such dynamic changes in attachment relations with the lost person could be critical to understanding the development of persistent and severe grief reactions, yet are not captured by attachment style measures (cf.Eisma, Tõnus, & de Jong, 2022).The fact that we found differences in the cross-sectional associations between attachment avoidance and prolonged grief symptoms dependent on whether people report general attachment tendencies or attachment to the deceased illustrates the potential importance of studying specific attachment relationships.Therefore, future multi-wave longitudinal survey studies should focus on how temporal changes in attachment relations relate to changes in grief severity.
Another issue that may partially explain our findings is that persons high on attachment anxiety may overreport their emotional responses to loss to gain social support from others.Anxiously attached individuals could subsequently benefit more from support if it is provided; a systematic review showed that people high on attachment anxiety are more engaged with treatment and more likely to seek out treatment (Adams et al., 2018).This potentially explains why attachment anxiety relates concurrently to higher prolonged grief symptom levels, but is not implicated in temporally increasing symptom levels.Moreover, it can explain why attachment anxiety may in some circumstances decrease prolonged grief symptoms (Wijngaards-de Meij et al., 2007a) and predict more treatment gains in internet-based prolonged grief therapy (Schmidt et al., 2022).Conversely, attachment avoidance is related to lower treatment engagement and participation (Adams et al., 2018), so it is perhaps unsurprising that it does not show similar effects as attachment anxiety.
Further adding to the complexity to the interpretation of our findings, is that attachment styles can change over time due to major lifeevents (Fraley et al., 2021), including interpersonal losses (e.g., Davila & Sargent, 2003).We identified no prospective studies in which adult attachment before bereavement was used to predict changes in prolonged grief symptoms after bereavement.It could be that changes in attachment styles occurring due to the experience of loss make it more difficult to detect temporal relationships between attachment style and prolonged grief symptoms assessed post-loss.It may also be that attachment styles predict the development of high acute grief, but do not lead to the persistence of grief.We cannot rule out these possibilities.

Limitations
Some limitations of our database are relevant to consider.First, there was a predominance of cross-sectional studies conducted post-loss.We recommend conducting well-designed large-scale multi-wave longitudinal studies, preferably including pre-loss and post-loss measurements of attachment styles and prolonged grief.Second, study samples predominantly included female, middle-aged bereaved adults.While this is common in bereavement research (Eisma & Stroebe, 2021), it potentially limits the generalizability of findings.Third, adult attachment was frequently operationalized along dimensions of attachment anxiety and attachment avoidance while a limited number of investigations directly examined persons classified with specific attachment orientations.Fourth, all studies used self-report measures of attachment styles.Using different methods of assessing adult attachment, such as the Adult Attachment Interview (George et al., 1985), may offer a more finegrained understanding of the relationships under investigation.Fifth, most studies included older measures of prolonged grief symptoms that do not correspond with current conceptualizations of PGD in the ICD-11 and DSM-5-TR (Treml et al., 2020;Lenferink et al., 2022).In part as a consequence, no studies provided information on the diagnostic status of participants regarding these new conceptualizations of pathological grief.
The methods used in our systematic review also come with limitations.First, we only considered English language peer-reviewed literature.We may have missed relevant studies published in other languages or in grey literature, such as dissertations.Second, our review only included one bereavement outcome, prolonged grief symptoms.We could therefore not clarify whether adult attachment is adaptive or maladaptive in general but we could only assess its relevance to understanding prolonged grief severity.Third, our meta-analyses were limited to cross-sectional associations, due to a large variability in longitudinal designs and analyses, which precluded quantitative synthesis of findings from those designs.

Conclusion
Notwithstanding the above-mentioned considerations, our M.C.Eisma et al. systematic review provides a timely and comprehensive overview of quantitative research on the associations between adult attachment and prolonged grief symptoms.Main findings are that there are small positive concurrent and longitudinal associations between attachment anxiety and attachment avoidance and prolonged grief symptoms.However, attachment styles did not predict changes in prolonged grief symptoms.We also found preliminary indications that in some circumstances people higher on attachment anxiety could show stronger reductions in prolonged grief severity.Grief experts should reconsider the current central role of adult attachment styles in prolonged grief theory pending further empirical investigation.Future multi-wave longitudinal studies, including pre-and post-loss measures, are recommended to elucidate how changes in attachment relations with the deceased relate to the development and persistence of prolonged grief.Ultimately, we hope this will help us improve understanding of how adult attachment affects our responses to bereavement and to improve care for severely distressed bereaved persons.

Funding
Maarten C. Eisma was supported by a Dutch Research Council (NWO) Veni grant [Grant ID: 016.veni195.113].The funder did not play a role in the study design, collection, analysis or interpretation of the data, in the writing of the report, or in the decision to submit the article for publication.

Declaration of competing interest
None.

Fig. 4 .
Fig. 4. Subgroup analysis comparing concurrent associations between attachment avoidance and prolonged grief symptoms for studies measuring attachment avoidance in general vs attachment avoidance in relation to the deceased.

Table 1
Summary of main findings on adult attachment and prolonged grief symptoms.
Table 2 summarizes study quality assessments.Sample Note.AAS: Adult Attachment Scale; ANQ: Attachment Network Questionnaire; ECR: Experiences in Close Relationships Questionnaire; ECR-R: Experiences in Close Relationships Questionnaire -Revised; ECR-RS: Experiences in Close Relationships -Relationship Structures Questionnaire; ECR-S: Experiences in Close Relationships Scale-Short Form; ICG: Inventory of Complicated Grief; ICG-R: Inventory of Complicated Grief -Revised; ICGS: Inventory of Complicated Grief Screen; PG = Prolonged grief; PG-13: Prolonged Grief Disorder-13; PGD = Prolonged grief disorder; RAAS: Revised Adult Attachment Scale; RSQ: Relationships Scales Questionnaire; RQ: Relationship Questionnaire; T1 = first time-point of measurement; T2 = second time-point of measurement.

Table 2
Quality assessment of included studies.

Table 3
Attachment anxiety and prolonged grief symptoms sensitivity analysis.

Table 4
Attachment avoidance and prolonged grief symptoms sensitivity analysis.